Bug #69187 [Asn]: json_last_error return BC in PHP7
| From: | bukka@php.net | Date: | Fri, 06 Mar 2015 12:53:18 +0000 |
| Subject: | Bug #69187 [Asn]: json_last_error return BC in PHP7 | ||
| References: | 1 | Groups: | php.bugs |
| Request: | Send a blank email to php-bugs+get-191213@lists.php.net to get a copy of this message | ||
Edit report at https://bugs.php.net/bug.php?id=69187&edit=1
ID: 69187
Updated by: bukka@php.net
Reported by: hdtfonseca at gmail dot com
Summary: json_last_error return BC in PHP7
Status: Assigned
Type: Bug
Package: JSON related
Operating System: linux
PHP Version: master-Git-2015-03-04 (Git)
Assigned To: bukka
Block user comment: N
Private report: N
New Comment:
@nikic I think you are right. It should stay as a standard string conversion and we shouldn't
have a special "hack" exceptions for specific types (null, false). I also think that we
shouldn't have error for true because then we would have to have error for scalar int or float
which would be unnecessary BC break. As such I think that it will be best to add just note about the
conversion to the UPGRADING where is already mentioned that empty string will emit error.
btw. The decoding JSON values that appear outside of object or array literals is not just a PHP
extension since March 2014 when RFC 7159 became a standard (see https://tools.ietf.org/html/rfc7159#section-2
) ;)
Previous Comments:
------------------------------------------------------------------------
[2015-03-06 12:20:18] nikic@php.net
@hdtfonseca: PHP also supports (as an extension to JSON - this is not supported by the standard
itself) decoding JSON values that appear outside of object or array literals. E.g. it's
possible to decode the string "1" to the value int(1). It just so happens that the string
representation of bool(true) in PHP is "1", which is why json_decode(true) === 1 (without
error).
------------------------------------------------------------------------
[2015-03-06 11:37:25] hdtfonseca at gmail dot com
I think @nikic has a point. What about, TRUE, 1 and 0 values? Shouldn't they all output
JSON_ERROR_SYNTAX?
In PHP 5.6 they all return JSON_ERROR_NONE but I think that 1, 0, "", TRUE, FALSE and NULL
should be consistent. At least "", FALSE and 0 should.
------------------------------------------------------------------------
[2015-03-06 07:23:08] nikic@php.net
@bukka I think the current PHP 7 behavior is correct. The json_decode function accepts a string and
is documented to accept a string. According to our coercion rules null and false correspond to the
empty string "", which is rightly no longer valid.
As such I don't think anything needs to change here. json_decode does not give you any
guarantees that json_decode($nonString) === $nonString and I don't see why we'd want to
introduce such a behavior.
------------------------------------------------------------------------
[2015-03-05 19:23:14] aharvey@php.net
Well, that was embarrassing. :)
Thanks for looking into this and in advance for the patch!
------------------------------------------------------------------------
[2015-03-05 18:23:46] bukka@php.net
Hi Adam, this is actually caused by your commit ( https://github.com/php/php-src/commit/a7b3abe4e6f5e2fdfd8d55b676c9ca6b3f9c8cc8
) :). The reason is the null is converted to empty string... The same will be for false.
I think that it makes sense to not emit error for null or false but only for empty string.
json_decode already returns number when you pass number (json_decode(1)) so it makes sense to have
it consistent. There is still a small inconsistency with boolean - json_decode(false) returns NULL
and json_decode(true) returns 1. However I'm not sure if it's worthy it to break BC (even
slightly)...
I will apply the patch for NULL and FALSE in a couple days if no objections.
------------------------------------------------------------------------
The remainder of the comments for this report are too long. To view
the rest of the comments, please view the bug report online at
https://bugs.php.net/bug.php?id=69187
--
Edit this bug report at https://bugs.php.net/bug.php?id=69187&edit=1